Originally published at: https://developer.nvidia.com/blog/advanced-ai-and-retrieval-augmented-generation-for-code-development-in-high-performance-computing/
In the rapidly evolving field of software development, AI tools such as chatbots and GitHub Copilot have significantly transformed how developers write and manage code. These tools, built on large language models (LLMs), enhance productivity by automating routine coding tasks. Parallel computing challenges However, the use of LLMs in generating parallel computing code—essential for high-performance…
If I understand this Advanced RAG solution, it recommends code for HPC development that is more syntactically accurate and operationally efficient than previous AI models AND it does so without requiring HPC-specific fine-tuning. Previous LLMs generate serial code effectively but struggle with parallel operations such as deadlocks and race-conditions. They also do not account for user code running efficiently on diverse HPC architectures with unique hardware complexities.
Sandia’s contribution to the solution seems to be some level of automation and integration of Kokkos which is a leading tool for abstracting performance-portable applications from the underlying hardware.
Also Sandia’s benchmarked improvements in query relevancy, accuracy and balance of breadth vs. depth are very promising. Please confirm.
You captured the overall messages from the RAG blog and Kokkos portability. If you wish to discuss further, I recommend to reach out to Sarah Tsai, co-author at Sandia, and implementer of the RAG model described. You may find her on LinkedIn too.